Marcus Johnson (00:04):

Hey, gang. It's Friday, January 9th. Nate, Jacob, and listeners, welcome in to Behind the Numbers, an EMARKETER video podcast. I'm Marcus, and joining me for today's conversation we have two gentlemen. The first is our principal AI analyst living in New York, it's Nate Elliott.

Nate Elliott (00:20):

Hey there, Marcus.

Marcus Johnson (00:21):

Hello, fella, live from the studio, and we're also joined by our senior analyst covering artificial intelligence and technology living in California. It's Jacob Bourne.

Jacob Bourne (00:30):

Hey, Marcus.

Marcus Johnson (00:31):

Hello, sir. We start, of course, with the Fact of the Day. So it doesn't seem that far away. Maybe it does. Not to me, but if it would take you about a year and a half to drive to the moon, if you were thinking about doing that. It's about 240,000 miles away. So if you're driving about 60 miles an hour, reasonable pace, for about eight hours a day, it would take you 18 months. If you didn't stop, you would get there in about five and a half.

Jacob Bourne (01:02):

I think that 60 mile per hour speed limit seems pretty low for a trip to the moon.

Nate Elliott (01:08):

That not a lot of traffic in the outer atmosphere.

Jacob Bourne (01:11):

It's got to at least be 75.

Marcus Johnson (01:13):

All right. We'll crank it up.

Nate Elliott (01:15):

The first time I sailed across the ocean, one of the deck officers... Because you're out there at the middle of the night without not a lot to do other than make sure you don't hit things, and so you have some pretty wild conversations, and one of the deck officers would keep bringing up the notion that if we drained the ocean, you could drive from Britain to America and vice versa, which is not quite as far as the moon, but might take as long if the bottom of the ocean bed is all mucky.

Marcus Johnson (01:45):

Who suggested this, the guy saying the-

Nate Elliott (01:48):

... on the STS Tenacious. Yeah.

Marcus Johnson (01:51):

Well, he's out of a job.

Nate Elliott (01:53):

Yes, that would've put him out of a job, but he would've gotten to drive from Britain to America and stop and camp along the way, which he seemed pretty excited about.

Jacob Bourne (02:01):

It's not the most practical of ideas, but it makes for a nice visual anyway, I guess.

Marcus Johnson (02:06):

You sailed across the Atlantic?

Nate Elliott (02:08):

Yeah, twice.

Marcus Johnson (02:10):

Good God. And how long did that take?

Nate Elliott (02:12):

Less than 18 months.

Marcus Johnson (02:13):

18 months? He's gone for four years, Marcus.

Nate Elliott (02:17):

That takes a couple months, but you're stopping along the way and sailboats are not the most efficient mode of travel anymore.

Marcus Johnson (02:23):

Where are you stopping?

Nate Elliott (02:24):

Islands mostly. Canary Islands if you take the southern route. Iceland, if you take the northern route.

Marcus Johnson (02:31):

Oh, okay. There's like a BP that's off to the side. It's quite a distance between earth and the moon. So you can fit every planet in our solar system between the two. It feels like that the moon's closer.

Jacob Bourne (02:46):

I'm going to have to fact check you-

Marcus Johnson (02:48):

Jacob never believes any of my facts of the day.

Nate Elliott (02:53):

It's not the Fact of the Day, it's the Not Possible Fact of the Day. Jacob thinks that Marcus is having a hallucination.

Marcus Johnson (03:01):

I spend way too long on these as Stewart knows, and so it has to be true. It might not be. Anyway, today's real topic. The great Behind the Numbers Bake Off/Take Off GenAI Trends to Watch in 2026.

(03:17):

All right. In today's episode, our takers or bakers will be cooking up some GenAI trends to watch in 2026. Takers because they're giving takes. Very clever, Marcus. Three rounds. One, Signature Take; two, the How It Will Technically Play Out Challenge; and three, the Show-Stopping Argument. Let's get into it. Round one. So in round one, our chefs will have one minute to explain the premise of their trend. We'll go back and forth between the two. Nate, you're up. What will you be cooking up for us today?

Nate Elliott (03:49):

Wait, am I a chef, a taker, a baker, or a cook?

Marcus Johnson (03:53):

All of them.

Nate Elliott (03:53):

Okay. Well, in that case, my take/bake is that by the end of 2026, Google will have overtaken OpenAI as the leading source of consumer AI engagement. I think that even though OpenAI and ChatGPT have been the undisputed leaders in consumer AI for three years now, since ChatGPT first launched, that they will stumble in 2026 and that Google is already a lot closer to ChatGPT than most people think, and that this is the year that Google's tools, be they AI mode or Gemini or others are going to actually take the lead.

Marcus Johnson (04:36):

Mm-hmm. And so when you say closer for folks, what are you looking at that points to that trend?

Nate Elliott (04:45):

I'm looking at data from those companies. I'm looking at third-party data on number of visitors and users. I'm looking at app downloads. When I say that Google's going to take the lead, what I mean is more people will use Google's tools for AI by the end of 2026 than use OpenAI's tools for AI.

Marcus Johnson (05:04):

You have some numbers in your research showing that you say we estimate 800 to 900 people million people already use Gemini and AI mode each month versus the 800 million that OpenAI says use ChatGPT each week, and as you just mentioned, by the end of next year, you expect Gemini to have more folks using that.

Nate Elliott (05:26):

I think Google, not necessarily Gemini itself. I mean, one of the mistakes Google's made over the past few years is splitting up their branding. Somehow Microsoft has done a better job of branding its AI tools than Google has in what might be a first. At Microsoft, you've got Copilot, whether it's the summaries on the search results pages, whether it's the standalone AI app for your phone, whether it's the AI assistant within other Microsoft products, it's all branded Copilot. Google hasn't done that. They've got AI overviews and AI mode and Gemini and Nano Banana.

Marcus Johnson (06:04):

Started with Bard.

Nate Elliott (06:05):

Yeah, started with Bard. They've got a lot of different brand names, but when you add all these tools up, I think that Google is likely very close to one billion monthly users, and the reason I think they're not over one billion is I think they'd be shouting about that if they were, but I think they have very close to one billion monthly users. And yeah, one billion or so weekly users, which is probably about where ChatGPT is as we record this, is more than one billion monthly users, but that gap is closing pretty quickly.

(06:35):

And for at least the last six months, more people have downloaded the Gemini app to their phones than downloaded the ChatGPT app to their phones, and so the lead has been consistently narrowing for over a year. And I think with so many consumers still uncommitted to an AI platform, 80 or 90% of people don't use AI on a very regular basis still. They're still experimenting with it or not using it at all. There's a lot of open field for Google to take, and it seems very clear that they're really well-positioned to take advantage of that opportunity.

Marcus Johnson (07:12):

Very nice. Let's move to Jacob, your prediction for us trend for 2026 is?

Jacob Bourne (07:19):

Yeah, my bake is that the AI boom that we've seen over the past three years is headed for a bit of a reality check in 2026. Now, that doesn't mean the opposite, we're not going to see doom. So boom isn't going to turn into doom. However, I think that we're starting to see some our industry pain points emerge, these structural bottlenecks, physical limitations, financial risks are going to throttle the breakneck pace that we've seen for AI so far. And one of the core issues is just that the supply chain can't quite keep up with AI companies' needs. So the demand for AI is still very real, enthusiasm is still very high, especially among the enterprise, but there's just real world constraints like energy. AI is very energy-intensive. There's realistic timelines for the infrastructure build out that really is necessary for the expansion. There's various component shortages.

(08:20):

And then underlying all of this is just the economics of AI are still quite messy, and so I think that this is going to really pose some inconvenient constraints in 2026 for just AI's overall acceleration. So I guess to put it in a nutshell, there's a lot of resources that go into making the magic of AI happen, and a lot of those resources are just under a lot of strain right now, and so that strain is going to come into focus in 2026.

Marcus Johnson (08:51):

Yeah. All right, gents, let's move to round two. It's the How It Will Technically Play Out Challenge. Our chefs will explain in more detail how they expect the trend to manifest in 2026. So we go back to, Nate, to your take that Google will overtake open AI in 2026. How do you see this playing out? All

Nate Elliott (09:09):

Right. Well, first of all, I'll go back to the point that most people still don't use AI on a weekly or daily basis. Fewer than half of American online users are even going to generative AI tools on a monthly basis, according to our latest forecasts. If you zoom that into what we might call fuller adaptation to borrow a term from Jacob's reports, we'd be looking at people using these tools weekly or even daily, the way that they use Google and the other tools that are part of their lives weekly and/or daily, and the vast majority of the population isn't there yet. 80 to 90% of online users in this country are not at that point with AI. So there's all this opportunity for users left to claim and no one's better positioned to claim those users than Google, in part because they have reach that certainly OpenAI can't match. Google offers the world's number one mobile operating system, the world's number one web browser, the world's number one search engine, the world's number one web mail platform, and on and on the list goes. No one can-

Marcus Johnson (10:16):

And they have about a billion people in each of those categories. It's not just that they're the leader, they have a billion people across all of those different-

Nate Elliott (10:22):

Three billion actually when you look at-

Marcus Johnson (10:23):

At least. At least.

Nate Elliott (10:25):

Yeah. So they've got probably a couple dozen products now with a billion or more monthly users. No one has the same ability to reach an enormous audience and convince them to use a new offering, and Google's been doing this very successfully for years. One of the reasons they have so many tools with a billion users is that they leveraged their other tools that already had a billion or more users to introduce these new tools. And even when they're late to market, remember, Android was not the first mobile smartphone system, Chrome was not the first browser, Gmail was certainly not the first webmail system. They were able to leverage their search leadership, their search dominance, to successfully launch and build into leadership all these other tools, and so it seems a really smart bet that they would have a good chance of doing that in the world of AI as well.

(11:18):

When you add to that, the fact that Gemini actually appears to be better for consumer use cases than ChatGPT, then it starts to look like Google has a really strong opportunity here, and I say that not because of the launch of Gemini 3. Even before Gemini 3 launched, if you looked at blind taste tests, if you looked at sites where people can go and compare two anonymized AI responses side by side and choose the one that they prefer, Google's tools consistently outperformed OpenAI's tools for almost every consumer AI application, whether it's search within AI, whether it's writing text and other content, whether it's creating and editing images and videos, Google's tools have been outperforming OpenAI's tools. And so when you add those two things up, it just makes sense that as more people start to explore AI, as more people build from experimentation into adaptation, that Google's going to have buyed into the fact that it has better tools and more reach and therefore marketing opportunity, Google's going to have the opportunity to overtake OpenAI and ChatGPT, and that's what we've been seeing play out for the last year.

(12:32):

We've seen global AI share data showing that in the last 12 to 15 months, OpenAI's share of global AI usage has gone down by about 15 percentage points. Now, that's from like 87% down to 72%. So they're still clearly in the lead and yet that lead is slipping very quickly. At the same time, you've got more people downloading Gemini than are downloading ChatGPT, and that's one of the reasons that share is changing so quickly. And I think 2026 is a year that that will finally tip and Google will in fact take the lead in consumer AI usage.

Jacob Bourne (13:10):

Yeah. Yeah, I think that Google's really its biggest advantage on this front is its reach. I mean, even if someone is going to Google search and doesn't intend to use AI, it's going to increasingly encounter AI when Google makes AI mode a more central search element, and so I think that alone just gives Google such a huge advantage. And then of course, you also look at just Google's vertical integration of AI. I mean, there's search, there's the big data element, there's its historical research innovation on AI. Then of course you have the cloud, it's getting revenue through AI cloud customers. Plus it has chips too, it has infrastructure. It's now supplying players like Anthropic and Meta with AI chips. So it just has this very robust and very integrated vertical AI stack that makes it very competitive, and of course on the consumer end, it's the reach that I think will make it sort of the default AI provider.

Marcus Johnson (14:20):

So could one of the concerns there be though... There was a few pieces I was reading for this and one of them was saying that an awkward wrinkle for Google is that DOJ is attacking Google's distribution strategies of saying that that could make attempts to bundle Gemini tightly into search and Android trickier. What do we think?

Nate Elliott (14:37):

It's possible. I don't think any of that's going to be settled anytime soon. That seems destined to drag out for many years to come. Even if you peeled apart every individual product that Google offers and AI was just tacked onto search, the fact is they have more than three billion global monthly search users. Something like 95% of US online users visit Google search every month and the vast majority visit it every day that they're online, and if they wanted to just using search, they could become the AI leader tomorrow. If they flipped the switch and started defaulting people into AI mode rather than those traditional 10 blue links, they would have well over one billion monthly search users. They could have as many as three billion monthly AI users just by flipping that switch tomorrow.

(15:28):

Now, I'm certain they're not going to do that. One of the things they're really good at is ramping people into alternative search results slowly but surely, and then using that single search box that people love to use, whether it's at the top of the browser or on google.com, using that single search box to identify what kind of search is someone doing. Are they looking for news? We'll default them into news search results. Are they looking for something to buy? We'll default them into shopping search results.

(15:58):

And I think starting probably in Q1, we'll start to see screenshots of people who get defaulted into AI mode, even though they didn't do an AI search, they just did a general Google search. And I think by the end of 2026, maybe as many as 10 or even 15% of regular Google searches, the longer ones, the more complex ones are going to get defaulted into AI mode as the search results. I don't know it'll go much higher than that because the vast majority of searches can still be served really well with 10 blue links or shopping results or news results or map results or whatever else Google has to offer, but that is one of the ways in which Google will start to increasingly claim its share of the consumer AI space.

Jacob Bourne (16:40):

Yeah. And Marcus, just to add regarding your original question about the regulation, I mean, a common theme for tech regulation is it has been slow. I mean, the government just moves a lot slower than the tech industry. So even if they were down the road to implement some sort of constraint against Google's dominance in this space, Google would already have gained those users, gained the revenue, the data associated with all of this, and it just further gives it at an advantage for whatever it will do next. So that's been a common theme for why regulation hasn't been so effective for tech.

Marcus Johnson (17:21):

It is so interesting because we're talking about Google overtaking OpenAI, and in the Fortune piece I was reading, they were saying, "Perception's a funny thing. It can change quickly. Just a few years back, Google was widely perceived to be a tech behemoth with too much power. Then OpenAI came along and there was one Twitter user or X user who rightly noted of Altman's AI company, they somehow made Google look like the underdog good guys that everyone cheers for." But you're right, distribution's a big piece of this. Alistair Barr of Business Insider, who's a sister company of ours, was writing that who ultimately wins will depend a lot on distribution and less on the quality of the tech, which is pretty similar these days. Ajay Agrawal, a researcher on economics of AI saying people are shifting over to Gemini, not just because it's got a better model, but they're realizing that this capability is baked into everything, Nate, to your point at the beginning. Folks realizing, "Oh, it's already on my phone. Why wouldn't I just use this one?"

Nate Elliott (18:19):

It's already on my phone, but also it already integrates seamlessly with lots of the tools that I already use. You can have Gemini put things into Gmail for you. My wife and I share a grocery shopping list on Google Keep, and when one of us comes up with a recipe we'd like to make for dinner using Gemini, we can tell it to add those items to the grocery list on Keep, and it knows how to do that. These are the kinds of integrations that Google either already has or can very easily flip the switch on and that OpenAI will have to negotiate partnerships and build external technology connections and all the other things that are required to make that happen so that, yeah, even if Gemini wasn't as good as ChatGPT, those integrations would be a big advantage for them despite the quality of the tool. But it turns out again that in blind taste tests, people tend to prefer Gemini over ChatGPT for most consumer AI use cases.

Marcus Johnson (19:15):

Mm-hmm. Jacob, let's move to yours in the second round. Tell us more about the AI boom and how, as you say, reality will give it a bit of a squeeze in 2026.

Jacob Bourne (19:28):

Yeah. I mean, the AI boom is just what we've seen with the entire marketplace has just been so driven by the gains from these AI companies, most of them big tech companies, but as I argued originally, this growth, the speed, the pace has limitations, and it's really divided into two areas. One is just the physical limitations behind... Or the physical limitations of AI expansion. And the other is just that the money side of the equation for AI doesn't quite add up. So on the physical side, it's really electricity is one of the main bottlenecks. So electricity prices have surged about 267% in parts of the US with high concentration of data centers, and by 2028, what we might see a 20% power shortfall for data centers.

(20:24):

So this doesn't just mean higher consumer prices and consumer backlash against AI. It also means tangible effects for AI companies themselves. For example, Microsoft admitted that it literally has GPUs, AI chips, sitting idly because this doesn't have enough access to energy to power them. Now keep in mind that these GPUs are some of the most expensive elements for CapEx spending on AI, and so to have them just sitting idly because you don't have enough energy essentially translates into waste and also missed opportunity to scale, which is what these companies want to do.

(21:00):

The other thing we're seeing Oracle, which has made a major data center deal with OpenAI, actually had to delay that project to 2028 because of labor and material shortages. So these companies are racing each other, but part of that racing means building enormous and very expensive infrastructure, and they just can't do it on time. So there's these bottlenecks. Another issue is just the components. So the AI chips require memory components, and those memory components are used in a wide variety of technologies globally, and so now you're seeing data centers kind of snatch up all these memory chips, and what it means is price hikes for an array of consumer devices are coming down the road, but keep in mind that some of these consumer devices are next generation AI-enabled devices that tech companies are hoping is going to boost AI adoption. But if you have higher prices on these things because of these component shortages, then you're going to see constraints on adoption of next generation AI-enabled smartphones, wearables, and whatever else that they're going to try and pack AI into.

(22:09):

So that's kind of the physical limitation side of things. And then there's a financial side where you have a reality where AI models, the frontier models, are just very expensive to build and deploy commercially to the point that they're not profitable yet, and so to date, really, Nvidia has been one of the biggest financial beneficiaries of the AI boom because it provides these chips, which are basically the basic building blocks of AI. But the problem is that even NVIDIA is now starting to see the headwinds, and basically it's... One of the things that happened recently is that the quality of Nvidia's stock actually was downgraded, and the reason why is because even though it's still making a lot of money on these chips, it's profitable, it's taking longer to get paid, and it's because its customers are really these same AI companies that are facing all of these bottlenecks and these companies are increasingly taking on massive amounts of debt to expand their AI ambitions, which really involves buying Nvidia's chips, which then they have to take on debt to pay for. So Nvidia is getting slowed down payments, which is a concerning sign.

Marcus Johnson (23:18):

Yeah, I like this line. You have big tech's AI profits depend heavily on selling infrastructure to AI startups that are hemorrhaging cash.

Jacob Bourne (23:24):

Right, exactly. I also want to add too, that the chips themselves pose a hidden financial danger because they're one of the most expensive elements to the AI buildout, but their economic lifespan is actually uncertain. One estimate I saw puts their longevity at between four and six years. So if you think back to chips that were purchased in 2022, then by next year, I mean, some of those are looking at potentially becoming obsolete. And so on top of that, AI companies actually have been borrowing money using those very same chips as collateral. So if you then have those chips becoming depreciated on a faster timeline, then these companies are then left holding assets with diminished value all the while they're not even profitable yet, at least on the AI front.

Nate Elliott (24:13):

So when you said earlier that the economics are messy, what you actually meant is that they're potentially disastrously unstable?

Jacob Bourne (24:19):

They're potentially disastrous, yeah. I mean, I think next year... I don't anticipate disaster next year. I think the AI bubble... I anticipate well into the first half of the year, we're going to see a continued boom and it's going to be boosted by the Fed rate cuts, which we're seeing. Nvidia has about a half a billion dollar GPU backlog showing that there's still expansion that's happening, but those physical constraints, the financial risks, which eventually could spell disaster, I don't think it's going to happen next year, but they're piling up. And so I predict we're going to see some sort of market correction for the AI sector before the end of 2026, just based on all these issues.

Marcus Johnson (24:59):

Well, so Nate, you write the price for winning in AI is tremendous, but so the resources needed to compete. You say VCs' patience won't last forever in a market where there's no established ad model and only 3% of AI users pay subscription fees per Menlo Ventures. It's interesting because if you look at the space, it looks like things are going quite well, especially if you look at the stock market, a ton of money being invested, but when you look at money made, it's a different picture. So Nate, for you, if I was to say the AI boom will feel a reality squeeze in 2026 because of blank, what do you think is the main reason it's going to slow down next year?

Nate Elliott (25:40):

I'd have to agree with Jacob that the economics are unstable enough or messy enough that those economic headwinds are going to hit the investments that the AI platform companies need to keep growing and developing. I mean, the reality is this: We don't need 12 to 15 consumer-facing AI tools, and yet here we are with 12 to 15, probably more if you add up all the minor players, consumer-facing AI tools, and they all are competing for all the things Jacob was talking about. They're competing for the chips, for the compute, for the water, for the electricity, for all of these resources, and they're also competing for investment. I don't think these VCs and investors have limitless pockets.

(26:30):

I think the writing will be on the wall at some point in the next six to 12 months for a lot of these smaller players, and I think by the end of 2026, we're going to see several of them, including some that we think of as bigger names, but that still only have 1% or 2% of global AI usage share. We're going to see them have to find some M&A options to get acquired, to merge together, to do something, to combine resources either with each other or to get absorbed into a larger organization that has the pockets to make this a reality.

(27:03):

Because if you're Google, if you're Apple, if you're Microsoft or Facebook, you're generating hundreds of billions of dollars in revenue from other things in your business portfolio that can help those companies sustain these types of investments. But if you're one of these pre-public market companies that we all know the names of, you don't have hundreds of billions of dollars in ad revenues to help fund your AI efforts. You need to rely upon the continued confidence of your investors, and I can't see that all of them are going to maintain that continued confidence for another 12 months.

Marcus Johnson (27:38):

Yeah. All right, gents, that's the end of round two. Round three is the Show-Stopping Argument. So this is where our chefs will pull out their best closing arguments as to why their trend is most likely to happen. We go back to Nate's trend and he's talking about Google overtaking OpenAI in 2026. Take it away.

Nate Elliott (27:59):

All right. Well, this is Google's playbook. They wait for a market to get interesting enough that they want to play in it, they show up several years late to the market and within a few years they've taken over and they win. They did it with browsers where they were about 10 years late to market compared to Internet Explorer, and yet within three years of launching Chrome, it had won. They did it in email where Gmail didn't get its full public launch for about a decade after Hotmail and RocketMail and some of the early tools, and yet within a few years of launching Gmail to the public, they had won. They did the same thing with mobile operating systems and maps and almost everything else they've launched.

(28:39):

There is no one better at showing up just a little bit late to market and then building a really good tool that has really, really good distribution and winning the market. It's not so much a question of whether Google will overtake OpenAI in terms of consumer AI adoption, it's just a question of when, and based on all the data I've presented today, it looks pretty clear to me that the when will be by the end of 2026.

Marcus Johnson (29:05):

Really quickly, if you had to now go over to the other side of the courtroom and argue for ChatGPT for a second, what does ChatGPT, or sorry, OpenAI have in its arsenal to fight back next year? Is there one kind of main thing that they can lean on to help stem the tide of Google?

Nate Elliott (29:21):

I think the biggest thing they have in their favor is the zeitgeist. I mean, remember, even as the walls are starting to close in around them to a certain extent, they have the name recognition to continue to build in the consumer marketplace. And listen, they launched Sora 2 probably knowing internally that it would lead to flagrant copyright violations. We saw people illegally using Disney IP all over that platform within minutes of it launching, and the end of that story could have been, perhaps should have been, Disney suing OpenAI and taking a lot of money and/or a lot of shares in exchange for those IP violations, and instead, the end of the story was Disney partnered with OpenAI, invested a billion dollars and gave them a license. That's because of ChatGPT's name recognition and its status in the zeitgeist. And so if they can do a better job of continuing to leverage that name recognition, that existing position as the number one player, then that's their best fighting chance at staving off Google.

Marcus Johnson (30:29):

Yeah. All right, Jacob, the AI boom will feel reality's squeeze in 2026, your closing arguments?

Jacob Bourne (30:36):

Yeah. So just want to zoom out a little bit to explain why this messiness is happening. So back when ChatGPT launched in 2022, it really ignited this AI frenzy, promising trillions of dollars added to the global economy, which raised sort of anxiety about this winner's take-all scenario, and that's what really has sparked these three years of increasingly risky financial deal-making and unrealistic expectations. But some of these bets I think that we're talking about have been counterproductive, and it's not because AI isn't important or revolutionary, it's because these real-world limitations have really... They constrain how fast these investments can be put to meaningful use, which inherently means that there's some waste going on, and I think the even bigger downside to this horse race that we're seeing is that if we look back at AI's history, we see many decades of fits and starts in terms of investment.

(31:34):

Now, the current generative AI groundbreaking innovation that has kind of underpinned today's technology dates back to 2017, and that's when Google researchers published their transformer architecture research, which paved the way for ChatGPT. But the whole reason that happened is because... Or one of the main reasons why that happened is because back then, AI researchers really took a more academic approach and there was a lot of open sharing between teams and organizations. Now the AI sector has become a lot more insular, it's more competitive, it's kind of lost that collaborative approach, and there's reasons people might argue why that's a good and necessary thing, but I think the biggest downside there is that it actually makes breakthroughs much harder because you don't have that kind of crowdsourcing of methods and ideas.

(32:24):

And so there's all this money going towards scaling individual companies with their infrastructure, but I think what's being lost is that innovative collaboration at the model architecture level, which is really how we got here in the first place. So if you combine that lack of collaboration leading to slower innovation, and you combine it with the financial risks that we talk about and the structural limitations, ultimately you get setbacks in terms of AI development. And so I don't think we're headed towards another AI winter like we saw in past decades, but I think what's happening now is that there are real reality checks taking place on these kind of overheated investments that I think are going to really hit a bit of a wall next year.

Marcus Johnson (33:11):

Gents, very compelling arguments. End of round three, and of course, now I have to crown a champion and today's winner is... Drum roll going on in the background. Thank you for the Brooks and team for putting it in. Both of you win today. I can't pick. It's a draw.

Jacob Bourne (33:32):

Is that allowed in the bake off?

Marcus Johnson (33:34):

I'm back in England. It's land of soccer and we do that here for some reason. So yes.

Nate Elliott (33:40):

Five days of test cricket and no winner.

Marcus Johnson (33:44):

Twice a year in the NFL, they're like, "We're going to have a draw. We're going to for no reason." You both won because they were very compelling. I think both are absolutely going to happen. So brilliant trends and even more in both of their pieces of research that these were drawn from. If you want more GenAI or technology trends, Pro+ subscribers can head to EMARKETER.com, of course, and search for GenAI trends to watch in 2026 or tech trends to watch in 2026. The links to both reports are also in the show notes, but that's all we have time for for today's episode. Thank you so much to my guests. Thank you first to Nate.

Nate Elliott (34:19):

Thank you.

Marcus Johnson (34:19):

And thank you, Jacob.

Jacob Bourne (34:20):

Glad to be here, Marcus.

Marcus Johnson (34:21):

And of course to the whole production crew and to everyone for listening to Behind the Numbers, and EMARKETER video podcast. Hope to see you on Monday for more Behind the Numbers. Until then, happiest of weekends.